I loved the article. It reflects the sentiments that I've had myself about this for quite some time...
Parts of the advances in the use of convolutional neural networks for pattern recognition have been impressive in their own right, I think.
But, there's leap of enormous proportions from advanced pattern recognition to actual simulated intelligence. The two are on different planets altogether, so to speak, which is what most media articles seem to miss out on when they cover this topic.
It's good to finally see at least one article which puts this thing right. :)